Last active
August 26, 2020 14:50
-
-
Save slavailn/e820c72ecdf03107569a266b280ac443 to your computer and use it in GitHub Desktop.
Cluster sets of genes based on Jaccard distance (mostly stolen from Biostars post)
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
# test Jaccard | |
# Create a list gene sets | |
char1 <- c("gene1", "gene2", "gene5", "gene9", "gene10") | |
char2 <- c("gene2", "gene3", "gene5", "gene7", "gene10") | |
char3 <- c("gene7", "gene9", "gene10", "gene11", "gene12", "gene1") | |
lst <- list(char1, char2, char3) | |
# Function to calculate Jaccard distance | |
siml <- function(x,y) { | |
1-length(intersect(lst[[x]],lst[[y]]))/length(union(lst[[x]],lst[[y]])) | |
} | |
# get a data frame of all factor combinations | |
z <- expand.grid(x=1:length(lst), y=1:length(lst)) | |
# calculate Jaccard distances between each pair of gene sets | |
result <- mapply(siml,z$x,z$y) | |
# Turn into matrix | |
dim(result) <- c(length(lst),length(lst)) | |
# Convert to distance | |
dists <- as.dist(result) | |
# Cluster a plot a dendrogram | |
plot(hclust(dists)) | |
############ FUNCTION FOR GOstats on KEGG only #################################### | |
## The GOstats output has to have a last column with comma separated gene names | |
clusterJakkard <- function(res_table, plot_name) | |
{ | |
# res_table="Neuroblastoma_DOWN_SIG_GOstats_KEGG.txt" this has to be a file in my usual format | |
# plot_name="Neuroblastoma_DOWN_SIG_KEGG_clusters.tiff" | |
keggtab <- read.table(res_table, sep = "\t", header = T, quote = "\"", | |
stringsAsFactors = F) | |
keggtab$KEGGID <- paste("0", keggtab$KEGGID, sep = "") | |
keggtab <- data.frame(KEGGID=keggtab$KEGGID, Term=keggtab$Term, Genes=keggtab$Genes) | |
keggtab$KEGGID <- paste(keggtab$KEGGID, keggtab$Term, sep="::") | |
keggtab <- keggtab[,-2] | |
# Get the list of vectors | |
gene.list <- strsplit(as.character(keggtab$Genes), ";") | |
names(gene.list) <- keggtab$KEGGID | |
lst <- gene.list | |
siml <- function(x,y) { | |
1-length(intersect(lst[[x]],lst[[y]]))/length(union(lst[[x]],lst[[y]])) | |
} | |
z <- expand.grid(x=1:length(lst), y=1:length(lst)) | |
result <- mapply(siml,z$x,z$y) | |
dim(result) <- c(length(lst),length(lst)) | |
colnames(result) <- names(lst) | |
rownames(result) <- names(lst) | |
dists <- as.dist(result) | |
hc <- hclust(dists, method="ward.D2") | |
tiff(plot_name, width=500, height = 800) | |
par(cex = 0.8, mar=c(2,2,2,15)) | |
plot(as.dendrogram(hc), horiz=T, type = "triangle") | |
dev.off() | |
} |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
(mostly stolen from Biostars post) lol
Amen to that !